Selection of knee radiographs for trials of structure-modifying drugs in patients with knee osteoarthritis: A prospective, longitudinal study of lyon schuss knee radiographs with the definition of adequate alignment of the medial tibial plateau
The quality of medial tibial plateau (MTP) alignment, which is assessed by measuring the distance between the anterior and posterior margins (intermargin distance [IMD]) of the tibial plateau, and the reproducibility of alignment in serial radiographs are suggested to be key elements in determining the accuracy and sensitivity to change in knee radiographs in patients with tibiofemoral osteoarthritis (OA). We evaluated the influence of both MTP alignment and radiograph superimposition on the sensitivity to change in radiographic joint space narrowing (JSN) in knee OA.
The study group comprised 106 patients with knee pain (73 with OA). Lyon schuss radiographic images of the knee were obtained twice (at baseline [month 0] and 12 months later), using a standardized radiographic procedure. Computerized measurement of the IMD for the assessment of MTP alignment was compared with the grading of MTP alignment by 2 observers using a 5-point scale (excellent, good, fair, poor, bad). To obtain the rate of JSN, computerized measurement of the joint space width was performed at month 0 and month 12. The sensitivity of the joint space width to change over 1 year was evaluated by the standardized response mean (SRM).
The mean (±SD) IMD was 1.2 ± 0.9 mm. The correlation between scoring and computer measurement of MTP alignment was highly significant. The cutoff value for satisfactory alignment (excellent or good) was an IMD of ≤1.2 mm. In OA knees, the mean (±SD) annual rate of JSN and the SRM were statistically higher in knees with an IMD of ≤1.2 mm at both month 0 and month 12 (0.34 ± 0.50 mm and 0.68, respectively) than in knees with an IMD of >1.2 mm at month 0 and/or month 12.
The quality of MTP alignment at both baseline and the end point highly influences the sensitivity to change in radiographic JSN in knee OA. To obtain relevant data, only radiographs showing an IMD of ≤1.2 mm at both baseline and the end point would have to be analyzed in studies of structure-modifying OA drugs.
It is generally accepted that the measurement of joint space width (JSW) from standard radiographs remains the most simple tool available for measuring joint damage in knee osteoarthritis (OA) and consequently for evaluating the progression of cartilage destruction in trials of structure-modifying OA drugs (SMOADs) (1–3). In knee OA, joint space narrowing (JSN) is determined by longitudinal assessment of the minimum JSW of the medial tibiofemoral compartment on weight-bearing radiographs. Among the radiographic views of the OA knee, the most widely evaluated are the standard anteroposterior (AP) view in extension, the semiflexed AP view (4, 5), the semiflexed metatarsophalangeal view (6), the Lyon schuss view (7), and the fixed flexion posteroanterior (PA) view (8). The advantages and limitations of these protocols have been extensively discussed by Brandt et al (9).
Several longitudinal studies (5, 7, 9–21) have assessed radiographic progression of knee OA over periods of time ranging from 1 to 11 years. Published data show very wide variations of the annual rate of JSN, ranging from 0.03 to 0.60 mm/year. This leads to questions regarding the causes of these discrepancies, as well as the influence of both the radiographic procedure and the measurement method, particularly because most of the reports were abstracts in which details of the methodology were not provided. Among the studies of 1–3-year duration (which is the currently recommended duration for trials of SMOADs), the annual rate of JSN ranged from 0.06 to 0.14 mm/year (except in one study, in which the rate was 0.60 mm/year) when the standard AP view was used, 0.09 to 0.23 mm/year when the semiflexed AP view was used, 0.035 to 0.41 mm/year when the Lyon schuss view was used, 0.18 mm/year with the fixed flexion PA view, and 0.09 to 0.16 mm/year with the semiflexed metatarsophalangeal view.
In all likelihood, the large differences in the rate of progression of JSN observed in these studies may be attributable to methodologic problems. Indeed, good short-term reproducibility of a radiographic protocol is insufficient for predicting the quality of its long-term longitudinal performance (22). The quality of medial tibial plateau (MTP) alignment is supposed to be of importance for both the accuracy of JSW measurement and the sensitivity to change in serial radiographs over time (15, 20, 22). Alignment of the tibial plateau is usually assessed according to the method described by Buckland-Wright et al (5), which is based on the distance between the anterior and posterior margins (intermargin distance [IMD]) of the tibial plateau. MTP alignment is considered to be satisfactory when the IMD is ≤1 mm.
The aim of the present study was to investigate the validity of IMD assessment and the influence of both MTP alignment and the superimposability of paired radiographs on the apparent rate of JSN over a 1-year period of time, based on Lyon schuss radiographs, in patients with medial tibiofemoral OA.
PATIENTS AND METHODS
We selected 106 knee radiographs obtained from patients 40 years of age or older (mean ± SD age 60.9 ± 11.7 years) who were referred to a rheumatology outpatient department because of chronic (>3 months' duration) unilateral or bilateral knee pain. These patients had undergone radiographic examination twice, at baseline (month 0) and again 1 year later (month 12), according to the Lyon schuss protocol (20). Patient positioning for Lyon schuss radiographs was performed as previously described (20, 23) (Figure 1). The degree of flexion is set as a result of positioning the patient with the tips of the great toes, knees, thighs, and pelvis coplanar and in contact with the examination table. The source-to-film distance was 100 cm. The central x-ray beam was directed at the center of the joint, in the space between the tibial spines and the femoral notch. Fluoroscopy equipment was available in all centers, and the x-ray beam was then adjusted fluoroscopically to provide optimal alignment with the MTP. The radiographs were obtained in a variety of radiology departments by a variety of radiology technicians who, although they had been given instructions describing the technique, had not been trained specifically in its performance.
Grading the quality of MTP alignment.
The quality of alignment of the MTP with the x-ray beam was assessed by 2 experienced observers (EV and TC). Alignment of the MTP was graded as excellent, good, fair, poor, or bad, based on the experience of the observer regarding the apparent superimposition of the anterior and posterior rims of the plateau (5). Assessments were performed twice, 1 week apart, with the observers blinded to the names of the patients and the results of the previous evaluations.
Computer measurement of IMD.
The degree of alignment of the MTP with the x-ray beam was measured, based on the distance between the anterior and posterior rims of the plateau (24). Computerized measurement of the IMD was automatically performed at the intersection between the MTP and the vertical diameter of the femoral condyle, irrespective of the minimum JSW. The latter was automatically obtained from a circle fitted to the condyle from 3 points set with the mouse by a single experienced observer (Figure 2). Assessments of the month 0 and month 12 radiographs were performed by observers who were blinded to the names of the patients and the results of previous measurements.
Grading of OA.
The Kellgren/Lawrence (K/L) grade (25) was assessed at month 0 by a single observer. Knees with a K/L grade of >1 were classified as being osteoarthritic (n = 73 patients). The others (K/L grade ≤1) were considered to be healthy (controls).
Automated measurement of JSW.
Minimum JSW was measured by the same observer, using a novel version of digitized image analysis software (Holy β13; Actibase, Lyon, France) in which detection of the joint space contour was automatically performed by the computer with the help of an edge-based algorithm (24). The procedure for defining landmarks was as follows: all radiographs were oriented so that the medial compartment was situated on the right of the computer screen. The x-axis was defined by a vertical line obtained with a single click of the mouse on the nonosteophytic outer edge of the MTP. The vertical inner limit was then automatically delineated at a constant distance from the outer limit. The y-axis of the coordinate system was defined by a horizontal line obtained by a single click of the mouse on line A, just below the joint space. Automatically, a parallel line was drawn by the computer at a constant distance above line B. Within these landmarks, the bone contours of both the femoral condyle and the tibial plateau were automatically obtained (Figure 3). When the algorithm failed to delineate the contours, the reader could correct the computer-drawn contours as appropriate, by drawing them with the mouse.
The smallest detectable difference between 2 measurements of a single film was 0.12 mm. The intraclass coefficient of correlation between repeated measurements of the same film was 0.99, and the coefficient of variation (CV) was 1.15%. The CV was much lower in satisfactorily aligned MTP radiographs than in unsatisfactorily aligned radiographs (0.8% versus 3.5%) (24).
A computer database containing all measured data was created in the StatView 5.0 format (SAS Institute, Cary, NC). The interobserver and intraobserver reproducibility of the semiquantitative evaluation of MTP alignment was obtained using kappa statistics. A search for a correlation between discrete and continuous variables was made using logistic regression. Comparisons between groups were performed using Student's t-test, the Kruskal-Wallis test, or the Mann-Whitney test, as appropriate. P values less than 0.05 were considered significant. The sensitivity of the JSW to change over 1 year was evaluated by the standardized response mean (SRM).
One hundred six knee radiographs were assessed. At baseline, K/L grades 0, 1, 2, 3, and 4 were found in 10, 23, 53, 16, and 4 patients, respectively.
Semiquantitative assessment of the quality of tibial plateau alignment.
The interobserver and intraobserver reproducibility in judging the quality of MTP alignment (excellent/good versus fair/poor/bad) was excellent (κ = 0.99). The mean ± SD IMD was 1.2 ± 0.9 mm (range 0–4.6 mm). The correlation between observer scoring and computer measurement of MTP alignment was highly significant. The IMD never exceeded 1.2 mm in knees scored as satisfactory (excellent/good) (Table 1).
Table 1. Correlation between semiquantitative evaluation of MTP alignment and measurement of the IMD*
Except where indicated otherwise, values are the mean ± SD mm. MTP = medial tibial plateau; IMD = intermargin distance.
0.18 ± 0.41
0.84 ± 0.40
1.28 ± 0.25
1.94 ± 0.69
3.04 ± 0.77
Measurement of JSN.
Table 2 shows the rate of JSN according to the quality of MTP alignment. The mean ± SD JSW was 4.2 ± 1.1 mm (range 1.13–6.76 mm) at baseline and 3.99 ± 1.2 mm (range 0–6.80 mm) at month 12. The mean ± SD rate of JSN was 0.16 ± 0.05 mm/year (range −1.84 to 3.03 mm/year; SRM = 0.32).
Table 2. Rate of JSN according to quality of MTP alignment*
Patient group, MTP alignment at baseline/month 12
JSW at baseline, mm
Rate of JSN, mm/year
Values for joint space width (JSW) are the mean; values for joint space narrowing (JSN) are the mean ± SD. MTP = medial tibial plateau; SRM = standardized response mean; S = satisfactory; NS = not satisfactory; OA = osteoarthritis.
0.26 ± 0.45
S/NS or NS/S
0.07 ± 0.52
0.20 ± 0.55
Patients with OA
0.34 ± 0.50
S/NS or NS/S
0.11 ± 0.48
0.19 ± 0.49
0.05 ± 0.22
S/NS or NS/S or NS/NS
0.27 ± 0.72
In patients with an IMD of ≤1.2 mm on both the month 0 and month 12 radiographs, the mean ± SD annual rate of JSN was 0.26 ± 0.45 mm (range −0.46 to 1.7 mm), and the SRM was 0.58. In patients with an IMD of >1.2 mm at month 0 and month 12, the mean ± SD annual rate of JSN was 0.20 ± 0.55 mm (range −0.37 to 1.1 mm), and the SRM was 0.36. In patients with an IMD of >1.2 mm at month 0 only or month 12 only, the mean ± SD annual rate of JSN was 0.07 ± 0.52 mm (range −1.84 to 0.57 mm), and the SRM was 0.13. The difference between groups was significant (P = 0.04)
Patients with definite OA.
In patients with a K/L grade of >1 (definite OA), the mean ± SD JSW at baseline was 4.1 ± 1.1 mm. At month 12, the mean ± SD JSW was 3.9 ± 1.2 mm, and the mean ± SD rate of JSN was 0.19 ± 0.48 mm/year (SRM = 0.40). In patients with an IMD of ≤1.2 mm on both month 0 and month 12 radiographs, the annual rate of JSN was 0.34 ± 0.50 mm, and the SRM was 0.68. In patients with an IMD of >1.2 mm at month 0 and month 12, the annual rate of JSN was 0.19 ± 0.49 mm, and the SRM was 0.39. In patients with an IMD of >1.2 mm at month 0 only or month 12 only, the annual rate of JSN was 0.11 ± 0.48 mm, and the SRM was 0.23. The difference between groups was significant (P = 0.02). In contrast, no difference between groups was observed (JSN = 0.25, 0.29, and 0.20 mm, respectively, and SRM = 0.48, 0.50, and 0.48, respectively) when radiographs with fair alignment (IMD ≤1.6 mm) were considered as being satisfactory.
Patients without knee OA.
In knees with a K/L grade of ≤1 (no or doubtful OA), the mean ± SD JSW at baseline was 4.79 ± 0.8 mm (range 3.50–6.76 mm). At month 12, the mean ± SD JSW was 4.58 ± 0.9 mm, and the mean ± SD rate of JSN was 0.11 ± 0.46 mm/year (SRM = 0.23). In patients with an IMD of ≤1.2 mm on both month 0 and month 12 radiographs, the mean ± SD annual rate of JSN was 0.05 ± 0.22 mm, and the SRM was 0.23. In patients with an IMD of >1.2 mm at month 0 and/or month 12, the mean ± SD annual rate of JSN was 0.27 ± 0.72 mm (SRM = 0.37). The difference between groups was not statistically significant (P = 0.4)
These data confirm that large variations in the rate of JSN could be attributable to technical problems, particularly misalignment of the MTP, especially when misalignment is affecting only one of the paired radiographs (Figure 4). Indeed, when MTP alignment was satisfactory on both the month 0 and month 12 radiographs, the annual rate of JSN was 0.34 mm in OA knees and 0.05 mm in knees without OA. In contrast, the annual rates of JSN were 0.11 mm and 0.27 mm, respectively, when alignment was not satisfactory on one of the paired radiographs. The superimposability of paired images also appeared to be important, because both the annual rate of JSN (0.19 mm) and the SRM (0.39) were higher in knees with unsatisfactory alignment on both the month 0 and month 12 radiographs than in those with satisfactory alignment on only one of the paired radiographs (0.11 mm and 0.23, respectively). This implies that variability is greater if MTP alignment is not achieved in only one of a pair of radiographs than if it is not achieved in both. However, the present data show that MTP alignment is what is essential, because the rate of JSN (0.34 mm/year) and the SRM (0.68) were much higher in satisfactorily aligned paired radiographs than in those with poor alignment on both the baseline and end point films (0.19 mm/year and 0.39, respectively).
Standardization of the radiographic procedure and joint positioning, and consequently the quality of the radiograph, appears to be the key element for obtaining an accurate and reproducible JSW measurement (26–28). The JSW of the medial tibiofemoral compartment was demonstrated to vary with weight bearing, MTP alignment, x-ray beam inclination, rotation of the feet, and degree of knee flexion (26). The Lyon schuss view is a PA weight-bearing, fluoroscopically assisted radiograph of the knee in flexion (7, 20, 23). The resulting degree of knee flexion varies from 20° to 30°, depending on the patient, but remains constant in serial radiographs of a given subject, leading to high reproducibility of joint positioning. Fluoroscopy is then used to modify the x-ray beam angle in order to obtain good alignment of the anterior and posterior margins of the MTP. The optimal x-ray beam angle for obtaining satisfactory MTP alignment is ∼10–11° downward (20) but can vary widely between individuals because of large interindividual variations in the anatomy of the knee (23). Indeed, we previously observed that satisfactory MTP alignment could be achieved in only 60% of radiographs obtained without fluoroscopic control (versus 88% of those obtained using the fluoroscopically assisted protocol), when performed by specifically trained radiology technicians (23).
Other longitudinal studies published to date suggest that fluoroscopic positioning methods are superior to nonfluoroscopic methods with respect to reproducing the position of the knee in serial examinations (22). Fluoroscopic methods also appear to be superior with respect to achieving parallel alignment of the medial tibial plateau and the x-ray beam (15, 23, 24), a positioning marker that is strongly associated with sensitive detection of JSN in knee OA (22).
This study shows that an adequate protocol for radiographic measurement of knee JSW must fulfill 2 major conditions: good alignment of the MTP and good superimposability of successive radiographs. Buckland-Wright et al (5) proposed that the MTP should be considered in satisfactory alignment when the IMD of the tibial plateau is ≤1 mm, and the present data confirm this suggestion.
Furthermore, the quality of MTP alignment is supposed to be important in terms of the sensitivity to change on serial radiographs over time. In a prospective trial (20), the sensitivity to change in radiographic JSW was related to the quality of MTP alignment, as demonstrated by the SRMs, which were, respectively, 0.72 and 0.38 in satisfactorily and unsatisfactorily aligned radiographs. In the same study, an unexpected increase in JSW of >0.5 mm was observed in paired radiographs with unsatisfactory alignment but never in pairs with good alignment. This was also observed by Mazzuca et al (15), who showed, over a mean period of 2.6 years, that good alignment of the MTP in paired standard AP radiographs resulted in good homogeneity in the rate of JSN. In contrast, in knees with misalignment of the MTP on both of the paired radiographs, JSN was not detectable, and its standard deviation was 61% greater than that in knees with good MTP alignment. In OA knees with satisfactory alignment on both images, the mean rate of JSN was 0.26 mm/year and was significantly higher than that in OA knees with misalignment on one or both radiographs (15); these data are very similar to those in the present study.
Finally, it must be noted that standard radiography is associated with potential limitations in measuring OA changes over time, even if the validity of JSW measurements was demonstrated by Buckland-Wright et al, who reported a good correlation between the medial tibiofemoral JSW and the sum of the thicknesses of both the femoral and tibial articular cartilage as measured in double-contrast arthrography radiographs (28). With fully automated measurements, variations in x-ray penetration in analog radiographs and edge enhancement in digital radiographs resulted in inflated measurements of JSW. Furthermore, in some cases the increase in medial JSW could be explained by the development of significant lateral compartment narrowing (29). JSW in weight-bearing, extended-view radiographs of highly symptomatic OA knees can be modified significantly by variations in joint pain, and longitudinal variations in pain may confound changes in the radiographic thickness of the articular cartilage (27). It has also been suggested that extrusion of the meniscus could contribute significantly to JSN, even in the absence of articular cartilage thinning (30).
In summary, results of the present study suggest that: 1) the Buckland-Wright criteria, which are based on measurement of the IMD, for the assessment of MTP alignment are valid and must be systematically used to assess the quality of radiographs in trials of knee OA; 2) discrepancies between the results of different knee OA studies could be attributable, in large part, to methodologic problems; 3) the very low rate of JSN in recent studies of SMOADs (13, 18), in which no details were given on the quality of MTP alignment assessment, might be attributable to MTP misalignment on at least one of the paired radiographs; 4) in trials of SMOADs, it seems essential to carefully position the joint to be radiographed, to use fluoroscopy in order to obtain optimal MTP alignment, and then to apply exactly the same protocol to additional radiographs of the same patient; 5) ideally, to obtain relevant data, only radiographs with an IMD of ≤1.2 mm at both baseline and the end point would have to be analyzed in studies of SMOADs.